Quantitative evaluation of sand body connectivity in shallow water delta front based on extreme learning machine

 In order to better predict the connectivity of sand body in the shallow water delta front, taking the shallow water delta front sand body in the Kong 1st section of the Fenghuadian Oilfield in Huanghua Depression as an example, quantify and optimize the factors that affect the sand body connectivity, and then establish a nonlinear mapping model with the optimized parameters and the connectivity level of sand bodies to quantitatively predict the sand body connectivity by using extreme learning machine, and the predicted results are compared with those of support vector machine to analyze its effectiveness and advantages. Finally, apply the model to the study area to verify its practical application effect. The results show that: 1) Four parameters, including sand-ground ratio, interlayer thickness, interlayer density and permeability, were selected to predict sand body connectivity by using extreme learning machine algorithm, and the accuracy rate reached 92.33%. Extreme learning machine has higher training efficiency than support vector machine algorithm while ensuring prediction accuracy. 2) The extreme learning machine training model was applied to predict the sand body connectivity of ZV2-1 sand layer in the study area, and the dynamic verification coincidence rate is over 92%, which has a good application effect. 3) There are two types of inter-well sand body connectivity modes in the study area: lateral connectivity and internal connectivity. In the lateral connectivity mode, if the two sand bodies are laterally cut and have similar thicknesses, the connectivity is good. In the internal connectivity mode, if the interbeds within the sand body are not developed, the connectivity is good.